Dictate in · Obsidian
Obsidian has no dictation of its own, and the usual plugin route asks you to create an API key and pay your provider to type into your own notes. Dictate is a hotkey you hold. Speak, release, and the punctuated text lands at the cursor in the note you already had open.
Dictate is our app. Apple Dictation is free and works in Obsidian too, and this page says where each one stops.
Standup, and what it changed
Spoken, not typed
Pricing stays at one plan. Dana is rewriting the FAQ before Friday, and the changelog waits on that. Nobody wants the annual tier back.
Ask Dana for the FAQ draft
Pricing page notes
Prose lands in the note. Brackets stay on the keys.
First, what Obsidian has
Obsidian is a markdown editor over a folder of files. As of August 2026 it shipped no dictation, no transcription, and no voice control of any kind, in the desktop app or in its core plugins. Anything that turns speech into text in a note comes from somewhere else.
That is not a complaint. A note-taking app that stays a note-taking app is why people choose it. It does mean the question of how to dictate into it is answered outside Obsidian.
The three routes
Read these across. The setup line is the one that decides it for most people, because it is the part you do before you have written a single word.
Route 01 · free
Route 02 · your API key
Route 03 · $8/month
The plugin route is a real answer and some people prefer it: the audio goes to a provider you chose, and if you dictate rarely, a per-minute bill can come to less than a subscription. The cost it hides is administrative. You keep a key, you watch a meter, and the whole arrangement lives inside one app. If you want the do-it-yourself version of that trade in full, we wrote it up in OpenAI Whisper on a Mac.
The part nobody markets
Any page that promises you will speak an entire Obsidian note is selling you something. Saying “open square bracket open square bracket” is slower than pressing the key twice, and it litters the transcript with words you then have to delete. Voice is good at prose. It is bad at punctuation-shaped structure.
The workflow that survives contact with a real vault is a split one: type the skeleton, speak the meat. You will keep both hands on the keyboard for headings and links, and that is fine, because the keyboard was never the slow part of writing a note.
Speak this
Type this
Four moments
Vaults are mostly capture and mostly prose, which is the shape voice is good at.
01 · The daily note
Open the daily note, type the heading, hold the key, and say what happened. A paragraph you spoke in twenty seconds beats a bullet you did not write at all.
02 · Capture
Put the cursor in an inbox note and speak the thought whole, in the words you were thinking it in. Sorting it into the right folder is a later job, and a cheaper one.
03 · The recap
Dictate does not sit in calls. It transcribes live speech only, so the workflow is you, retelling the call in your own summary: what was decided, who owns it, what is still open.
04 · Literature notes
Paste the passage, then speak your own commentary underneath it. That second paragraph is the one that makes the note worth keeping, and it is the one people skip.
Under the hood
Dictate inserts through the macOS Accessibility API first. Where an editor is not a plain native text field, it falls back to paste, and if Accessibility access has not been granted at all it puts the transcript on the clipboard so nothing is lost. Nothing is installed into Obsidian, so nothing breaks when Obsidian updates.
The glossary is the part that matters in a vault full of proper nouns: canonical terms plus the way you actually pronounce them, reviewable and removable in Settings.
The trade
Transcription runs on our servers. Audio travels over TLS, is transcribed, then discarded; history stores transcript text and metadata, never recordings. Dictate is not local, and it needs a network connection to transcribe. Only the signed subscription ticket works offline.
If your vault is the kind that must never leave the machine, a local-model tool is the honest answer, and Apple Dictation on Apple silicon can process supported languages on device. We would rather say that here than have you find it out later.
Past Obsidian
The same held key works when you dictate into Notion for the document your vault note turns into, and the argument for holding a key rather than toggling one is set out in push to talk dictation on a Mac. If you have never dictated on a Mac at all, start with how to dictate on a Mac, and the rest of the app guides are in dictate in any Mac app.
Dictating into Obsidian
What people search before they talk into a note.
No. As of August 2026, Obsidian shipped no dictation of its own. Voice input into a note comes from macOS, from a community plugin, or from a system-wide app that types wherever the cursor is.
Click into a note so the cursor is blinking, hold the Dictate hotkey, say the sentence, and release. The punctuated text lands at the cursor. There is no plugin to install and no Obsidian setting to change.
It works, and installing it costs nothing, but it sends your audio to a Whisper-compatible transcription API using a key you create and a bill from whichever provider you picked, typically charged per minute of audio. Dictate is a flat $8 a month with no key to manage and no meter.
Yes, and it is free. Obsidian is a normal Mac app with normal text fields. Apple Dictation brings its usual limits: it stops after about 30 seconds without speech, automatic punctuation is hit and miss, and there is no custom dictionary.
Not usefully. Saying the hashes, brackets and pipes out loud is slower than typing them, and it reads badly in a transcript. The realistic workflow is voice for the prose and the keyboard for the structure.
Dictate has a glossary: canonical terms plus the spoken aliases you actually say, editable and removable in Settings. Add the project names and jargon your notes are full of. Transcription is English only.
Audio travels over TLS to our servers, is transcribed, then discarded. History stores transcript text and metadata, never recordings. Transcription needs a network connection; only the subscription check works offline.
Flat $8 a month, monthly only. Unused subscriptions are refunded in full; once you have transcribed something, that month is not refundable.
